
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
Dynamic Diuse Global Illumination using Light Field Probes 7
3 RESULTS
As demonstration, we render the modied Sponza scenario at 3840x2160 on a GeForce 1660ti laptop. Our implementation
is built on the Unity engine using the shaders from our supplement. Without the hardware raytracing acceleration, our
realtime rendering achieves a stable frame rate over 60fps. The render results could be nd in the appendix.
4 FUTURE WORK
Although the experiment result indicates a scalable and promising performance of the proposed approach, several
drawbacks and corresponding tentative future work are worth to note: (1) Inaccurate shadowing caused by the nature
of cascade shadow map: As the properties of the camera change in real time, the corresponding frustum will also
change, which may cause the shadows to change dramatically in two consecutive frames, resulting in shadow jitters
and lacking temporal coherence. In this paper we have taken the approach of sampling the highest level cascade to
guarantee maximum coverage to alleviate this problem, but this has led to the problem of shadow aliasing being more
pronounced. (2) Alternate irradiance update approach: the irradiance map is utilized in current implementation for a
lower bandwidth usage during sampling stage (one sample per read). The problem, however, is that on the one hand
the memory footprint is large and on the other hand it is troublesome for lighting translucent objects and volumetric
eects. [
3
] provides a feasible solution: they convolve octahedron radiance into spherical harmonic basis to extract the
main light direction for illuminating translucent objects and volumetric fog; and use the obtained spherical harmonic
coecients to generate the irradiance map to speed up the indirect lighting of opaque objects. (3) Not applicable to large
scene: During the rendering process in runtime, despite the fact that we are doing downsampling, the huge number
of probes will take up a lot of memory. Adaptive placement of probes rather than uniform placement would alleviate
this problem. (4) Lack of high frequency details: The low-frequency nature of diuse global illumination allows us
to reduce the probe resolution, but this also results in a lack of high-frequency details. The use of some screen space
techniques (e.g. screen space ambient occlusion, screen space global illumination) to supplement high frequency details
is a possible way to improve image quality.
5 CONTRIBUTION
William Xie conceived of the proposed idea. William Xie developed the oine processing tool. Brian Royston and
William Xie worked on the runtime probe update. Brandon Shin and William Xie worked on the runtime probe sampling.
Han Wang conducted the sanity test and organized the manuscript. All authors contributed to the review of manuscript.
REFERENCES
[1]
Marc Levoy and Pat Hanrahan. Light eld rendering. In Proceedings of the 23rd annual conference on Computer graphics and interactive techniques,
pages 31–42, 1996.
[2]
Morgan McGuire, Mike Mara, Derek Nowrouzezahrai, and David Luebke. Real-time global illumination using precomputed light eld probes. In
Proceedings of the 21st ACM SIGGRAPH Symposium on Interactive 3D Graphics and Games, pages 1–11, 2017.
[3] Bruce Woodard. Advanced graphics summit: Lifting the fog: Geometry amp; lighting in ’demon’s souls’, Jul 2021.
[4]
Zander Majercik, Jean-Philippe Guertin, Derek Nowrouzezahrai, and Morgan McGuire. Dynamic diuse global illumination with ray-traced
irradiance elds. Journal of Computer Graphics Techniques (JCGT), 8(2):1–30, June 2019.
[5] Thomas Engelhardt and Carsten Dachsbacher. Octahedron environment maps. In VMV, pages 383–388, 2008.
[6]
Fan Zhang, Hanqiu Sun, Leilei Xu, and Lee Kit Lun. Parallel-split shadow maps for large-scale virtual environments. In Proceedings of the 2006 ACM
international conference on Virtual reality continuum and its applications, pages 311–318, 2006.
Manuscript submitted to ACM